huggingface-jobs

Execute Python workloads and Docker containers on managed Hugging Face cloud infrastructure.

2|Updated Mar 29, 2026

Search Tags:#automation #gpu #inference #huggingface #compute #jobs

Ready to run

npx skills add https://github.com/NagyVikt/codex-plugins --skill huggingface-jobs-nagyvikt

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: huggingface-jobs
Source: https://github.com/NagyVikt/codex-plugins/tree/main/plugins/hugging-face/skills/jobs
Command: npx skills add https://github.com/NagyVikt/codex-plugins --skill huggingface-jobs-nagyvikt

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires datasets, transformers, vllm, huggingface-hub, torch, numpy, tqdm, scikit-learn, polars, ascii-graph, flashinfer-python, hf-xet, and includes scripts (resource) and references (resource) components.

What problem does it solve?

This skill removes the friction of setting up local GPU or TPU infrastructure by allowing you to execute Python workloads directly on managed Hugging Face cloud infrastructure.

Core Features & Use Cases

Managed Compute: Run data processing, batch inference, and model training on CPUs, GPUs, or TPUs without local hardware setup.
Workflow Automation: Schedule recurring tasks or trigger jobs via webhooks for automated data pipelines.
Use Case: Use this skill to run a batch inference job on 10,000 samples using a GPU-accelerated vLLM container, automatically persisting the results to a Hugging Face dataset repository.

Quick Start

Use the huggingface-jobs skill to run the generate-responses script on the specified input dataset using an a10g-large GPU flavor.

Frequently Asked Questions about huggingface-jobs

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I run batch inference on Hugging Face cloud infrastructure without local GPUs?▼

You can run batch inference on Hugging Face cloud infrastructure by submitting arbitrary Python workloads through the hf_jobs MCP tool. This skill provisions managed CPUs, GPUs, or TPUs and automatically persists the inference results to a Hugging Face dataset repository.

Can I use vLLM for batch inference tasks on managed Hugging Face compute?▼

Yes, you can execute batch inference tasks using a GPU-accelerated vLLM container on managed Hugging Face compute. The skill supports executing Docker containers to process large datasets efficiently without requiring local hardware setup.

How do I schedule recurring ML workloads and automate data pipelines on Hugging Face?▼

You can schedule recurring ML workloads and automate data pipelines by triggering jobs via webhooks or setting up scheduled compute tasks. The skill manages job submission, hardware selection, and secure secret management for Hub authentication.

Does Hugging Face Jobs support secure secret management for Hub authentication?▼

Yes, Hugging Face Jobs supports secure secret management for Hub authentication. This allows you to securely execute Python workloads and Docker containers on managed cloud infrastructure while protecting your access credentials.

What is the best way to generate synthetic data using Hugging Face cloud infrastructure?▼

The best way to generate synthetic data on Hugging Face cloud infrastructure is to submit a Python workload using this skill. It handles compute provisioning and automatically persists the generated outputs directly to a Hugging Face dataset repository.

Do I need local GPU hardware to run large-scale model training with Hugging Face Jobs?▼

No, you do not need local GPU hardware to run large-scale model training. This skill removes local infrastructure friction by allowing you to execute Python workloads directly on managed Hugging Face cloud infrastructure with hardware selection.